Fact-Sentence Candidate Selection on XALIGN 1.0 (golden annotated)
0.902MR ScoremT5
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| mT5Training Strategy=Transfer learning2022.02 | 0.902 | 0.831 | 0.841 | 0.886 | 0.845 | 0.851 | 0.751 | 0.785 | 0.837 | |
| mT5Training Strategy=Distant supervision2022.02 | 0.79 | 0.714 | 0.776 | 0.786 | 0.766 | 0.8 | 0.698 | 0.705 | 0.754 | |
| XLM-RobertaTraining Strategy=Distant supervision2022.02 | 0.781 | 0.69 | 0.765 | 0.739 | 0.765 | 0.785 | 0.669 | 0.724 | 0.74 | |
| XLM-RobertaTraining Strategy=Transfer learning2022.02 | 0.772 | 0.767 | 0.78 | 0.812 | 0.79 | 0.805 | 0.831 | 0.727 | 0.786 | |
| MuRILTraining Strategy=Distant supervision2022.02 | 0.763 | 0.684 | 0.74 | 0.755 | 0.705 | 0.785 | 0.624 | 0.677 | 0.717 | |
| Stage-1 + TF-IDFTraining Strategy=Baselines2022.02 | 0.75 | 0.685 | 0.693 | 0.718 | 0.737 | 0.701 | 0.787 | 0.647 | 0.715 | |
| MuRILTraining Strategy=Transfer learning2022.02 | 0.716 | 0.717 | 0.765 | 0.751 | 0.734 | 0.787 | 0.795 | 0.718 | 0.748 | |
| WITA-styleTraining Strategy=Baselines2022.02 | 0.507 | 0.574 | 0.517 | 0.459 | 0.602 | 0.5 | 0.535 | 0.53 | 0.528 | |
| KELM-styleTraining Strategy=Baselines2022.02 | 0.493 | 0.426 | 0.368 | 0.451 | 0.41 | 0.372 | 0.436 | 0.338 | 0.411 |